Digital Signal Processing Reference
In-Depth Information
Boss et al. [343] carried out an extensive analysis of the spectroscopic
features of cancerous in comparison benign ovarian cysts. The cystic fluid
samples were from 12 patients with malignancy and 23 with benign ovarian
fluid. Both one and twodimensional in vitro MRS analyses were performed.
There were many differences in metabolite concentrations distinguishing the
cancerous and benign cysts in that study [343]. For instance, the concentra
tions of isoleucine (1.02 ppm), valine (1.04 ppm), threonine (1.33 ppm), lactate
(1.41 ppm), alanine (1.51 ppm), lysine (1.67 ppm - 1.78 ppm), methionine
(2.13 ppm), glutamine (2.42 ppm - 2.52 ppm) as well as choline (3.19 ppm)
were all significantly higher in the malignant samples. However, no metabo
lites were identified that yielded complete distinction between malignant and
nonmalignant cyst fluid.
Even though there have been noteworthy results for ovarian cancer diagno
sis, major problems remain that hinder broader use of in vitro MRS in this
clinical area. Due to these problems, in vitro findings still cannot be con
sidered the “gold standard” with which MRS signals encoded in vivo from
the ovary could be compared. Mountford et al. [344] suggested that via
the statistical classification strategy (SCS) highly accurate distinction could
be made between malignant and benign tissue, based upon identification of
specific spectral regions of key diagnostic importance. For ovarian cancer,
2D MRS was particularly important because of overlapping resonances, and
was reported to “provide unequivocal assignment of resonances from chemi
cal species that contribute to the various pathological states defined during
(ovarian) tumor development and progression” (p. 3692). However, in his
comments on the SCS used in the abovedescribed in vitro MRS analyses,
Gluch [345] questions the suitability of this methodology for more complex
pathological entities, stating: “A classifier can more readily be developed when
the likelihood is high of belonging to a class of either 'yes' or 'no', but when
a tissue undergoes numerous stages in evolution from normal to malignant,
SCS shows no superiority over conventional pathology” (p. 467). Notably,
the high diagnostic accuracy was achieved by excluding the fuzzy samples.
In corroboration with the statement of Gluch [345], it should be pointed
out that even in the study by Boss et al. [343], which has, to date, yielded the
most extensive in vitro analysis of MRS signals from benign and cancerous
ovarian samples, the cited ranges for each of the metabolites were wide and
overlapping. As a clinician, Gluch [345] enumerates a number of key limita
tions of MRS as a diagnostic tool in oncology. These include lack of reliable in
vitro databases and lack of specific findings. For example, narrow lipid reso
nances can also be seen with necrosis, inflammation and other nonmalignant
processes. He emphasizes that di
culties arise not so much for large lesions
with suspicious imaging characteristics, but instead for small or in situ le
sions. It is precisely for these early stages of cancer that maximally sensitive
and specific diagnostic methods are most urgently sought. Putting this into
the realworld clinical perspective, Gluch [345] notes the “chance of missing
a cancer of the order of only 1% would translate into a significant medicole
Search WWH ::
Custom Search